Closed Welxiao closed 6 years ago
The num_channels
parameter is just something I use when I build the model to determine how many channels will be in the first convolution.
It is used in the function build_model
, here is a link.
num_channels = params.num_channels
bn_momentum = params.bn_momentum
channels = [num_channels, num_channels * 2]
for i, c in enumerate(channels):
with tf.variable_scope('block_{}'.format(i+1)):
out = tf.layers.conv2d(out, c, 3, padding='same')
if params.use_batch_norm:
out = tf.layers.batch_normalization(out, momentum=bn_momentum, training=is_training)
out = tf.nn.relu(out)
out = tf.layers.max_pooling2d(out, 2, 2)
I put it in the params.json
file so that I can easily change how big the model is. So if the model is overfitting with num_channels=64
I could modify it to num_channels=32
.
thanks for your answer
Excuse me, Sir. Thanks for you code of triplet loss, but i don't understand the 'num_channels=32' means in 'params.json'. In general, the channels of a picture are 1 or 3, 'num_channels=32' perplexes me. Could you help me , thanks much.